H.264 as MTI for rtcweb Jonathan Rosenberg Bo Burman

The Cisco Announcement Redux Open Source under BSD and binary module we distribute, we pay MPEG-LA Binary versions for Win, Mac, Linux, Android – community can contribute builds for others. Build tools open sourced, you can verify binary signature we distribute matches your build from source Minimal constraints for us to pay – we need to distribute; “About” recognition; must be possible for user to disable Cisco commits to support and pay barring unforseen changes in H.264 licensing environment

What can you use it for? Open source – anything. Its BSD. If you distribute it you might be subject to MPEG-LA licensing terms. Note first 100k are free. Binary Module – Not restricted to webRTC – Can work for an OS – e.g., Debian can pull the binary module on install of Linux. – Can work for server software too Binary module usable for all things under MPEG-LA type (a) license See cuments/AVC_TermsSummary.pdf cuments/AVC_TermsSummary.pdf

Ship Date Cisco will push first version of source into public repo before the end of the year and hopefully earlier Binary version available only after Jan1 due to MPEG-LA cap

The Factors for Consideration FactorWhy and for whomH.264VP8 Interop with installed user base Enabler for existing players to build web clients Enabler for B2C apps (customer support) – interop with installed video base in contact center

B2C Example: Talk to Investment Broker Jill goes to bank website, wants to speak with her investment broker Bob. Bob’s bank has deployed videophones to desktops, he takes call from there. The bank is not going to rip out and replace their existing video devices and softclients and contact center with something new. They want INTEROPERABILITY.

The Factors for Consideration FactorWhy and for whomH.264VP8 Availability of Experts and tools Important for larger players to build their own – most do that today Multiple Software Codebases Many implementations in software of H.264. Almost uncountable number of applications VP8 – 1 codebase in usage 6 apps listed – 4 from Google Why? Flexibility, maturity SDO StandardImportant for change control – particularly for those who implement.

Hardware Acceleration In-Market Chipsets – 100% of all chipsets listed in Strategy Analytics Handset Components technology market share have acceleration for H.264 encode and decode in hardware – covering almost all in-market smartphones – Only 4 chipsets in market have VP8 acceleration: Qualcomm Snapdragon 800 (Goog Nexus 5, LG G2, Samsung Galaxy Note 3, Nokia Lumia 1520, Sony Experia Z Ultra) Samsung Exynos 5420 (Samsung Galaxy Note 3) Rockchip RK3xxx (Android tablets in Asia) Nvidia Tegra 4 (Android tablets by Asus, HP, Toshiba) Software API access for realtime encode/decode is variable with upwards trend in availability

The Factors for Consideration FactorWhy and for whomH.264VP8 Hardware Acceleration Enabler for higher quality on mobile, of somewhat diminishing importance as CPU speeds increase

Performance Evaluation 640 x x x 720 H.264 CBP vs VP8H.264 CHP vs VP8 No rate control H.264 1% betterH % better With rate control H.264 1% betterH % better PSNR (dB) kbps

The Factors for Consideration FactorWhy and for whomH.264VP8 QualityMinimum bar necessary for usability – but this is the MTI and represents the lower bound. Widespread commercial deployment of H.264 CBP today indicates clear viability. High probabilty of most browsers going to H.264 High (same license terms as CBP) which outperforms VP8.

Looking At Distribution Holistically Assess Distribution Options for VP8 and H.264 Assess Patent Risks for H.264 and VP8 Perform Financial Risk/Impact Analysis Conclusion Focus needs to be browsers and mobile apps – those are key for success of webRTC

Patent Risk H.264VP8 Years in-market102 Revenue base to target by those seeking compensation Enormous – billions (Blu- Ray, countless commercial products) $0 (all existing VP8 apps are free) Breadth of target companies Enormous – consumer electronics, software, SaaS, all geos and market segment Only two – Google, Mozilla (no longer in Skype) Extent of patent analysis work Many lawyers and many years New Existing lawsuits from patent holders NoneNokia – in progress ConclusionLow RiskHigh Risk Factors for risk of new patent-holders emerging and suing:

Distribution of H.264 – its about options CategoryOptionsComments Operating Systems 1.All major mobile device/OS combos already pay and distribute H Cisco binary distribution model available for all OS’s OS distribution not required for success of webRTC on desktop (sorry Fedora, Debian) Browsers1.Distribute and pay your own way 2.Cisco binary distribution model available (Mozilla) Mobile Apps1.First 100k are free - distribute your own 2.Android covered by Cisco binary module 3.Distribute >100k and pay your own way 4.Rely on OS distribution – Android, hopefully IOS Very few IOS apps every see 100k downloads

Its about Risk/Impact Assessment RiskLikelihoodImpact 100k distros on IOS, AND No solution from Apple AND Cannot afford $0.20 per app LowMed New patent holders emerge for H.264 and demand unreasonable fees LowHigh Nokia conclusion results in inability to distribute VP8 at all MedHigh Nokia conclusion results in ability to distribute VP8 but at moderate cost Med New patent holders emerge for VP8 and demand unreasonable fees as a consequence of IETF selection of VP8 as MTI and subsequent deployments High H.264 VP8 Analysis points to H.264 as the better choice

Overall Analysis Results FactorH.264VP8Blocker and for Whom? Interop with Install BaseVP8 for existing players Availability of ExpertsNo Multiple CodebasesNo SDO StandardNo Hardware AccelerationNo QualityNo Financial RiskVP8 too high for large players. For small players, “its not free” is a complaint but objective risk analysis still points to H.264

Conclusion Selecting VP8 will turn away the existing players due to interop and financial risk and introduces a real financial risk for the smaller players, likely causing webRTC to fail to reach critical mass Selecting H.264 will enable the existing players and is objectively the lower financial risk option. Higher chance of success for WebRTC. Work is active on achieving RF status for H.264 on two fronts – MPEG WebVC (CBP) and MPEG-LA revisit